There are two ETL business machines deployed PowerCenter in the production environment, and the recent discovery cannot connect to the ETL service through the client.
The initial suspicion is that the powercenter is dead or the database has failed.
Logging on to the device found that the PowerCenter process exists, but the database (Oracle) is dead.
Switch Oracle Account entry, manual startup library, results start times wrong: ora-01157,ora-01110. Reading the paper indicates that the data file cannot be locked or recognized.
Exit Sqlplus, find the data file exists, guess the failure caused by the data file corruption, because it is noarchivelog (non-archive) mode, can not be restored through the log.
Recall the previous useful EXPDP export a backup, in order to restore ETL production as soon as possible, only drop the data file re-established.
The approximate process is as follows:
Lsnrctl Stopsqlplus/As Sysdbasql>shutdown immediate; SQL>startupMount;
Sql>alter database Archivelog; SQL>alter database DataFile'/oradata/xxx.dbf'offline drop; SQL>ALTER DATABASE open; SQL>drop tablespace xxx including contents and datafiles; SQL>create tablespace xxx ... datafile'/oradata/xxx.dbf' ...; SQL>EXIT;IMPDP \'sys/as sysdba\ ' tablespaces=xxx directory=data_pump_dir dumpfile=xxx.dump logfile=impdp-xxx.log content= Metadata_onlyIMPDP \'/As Sysdba\ ' Schemas=etl_user directory=data_pump_dir dumpfile=etl_user.dump logfile=impdp-etl_user.log content= Data_onlylsnrctl Start
After the table space metadata and user data are imported, the data is basically restored to the previous backup point.
The following again for the PowerCenter access to the details of the table to make adjustments, and later restart the next PowerCenter to see the operation.
If PowerCenter backup of the storage service in advance, then the backup is used to restore the configuration, then basically there is no problem.
However, the above operation is slightly rough (a friend pointed out that the danger is great), so please think twice before gentlemen operation, otherwise, the more extensive data loss is not worth the candle.
Conclusion: Thousand good million good, backup best.
Reference Documentation:
http://www.itpub.net/thread-27492-1-1.html
ORA-01157, remember an Oracle recovery process